Sampling small older populations:

Methods and challenges of a dementia prevalence study

NZSA2024

Claudia Rivera-Rodriguez, PhD

New Zealand changing population

Image 1

The need for a prevalence study

Image 1

Feasibility study (2020)

Image 1

  • Found that the sampling/doorknocking strategy was reasonable

  • We were able to train up multi-ethnic interviewers

  • Response rate at the door-knocking stage was 75% but at subsequent stages was about 25%

  • Demonstrated that a prevalence study was feasible in Māori, Chinese, Indian and Pākehā, not in Pacific populations.

IDEA Study

  • AIM: Establish the true current and future prevalence of dementia in NZ, & disparities
  • This study is for Chinese, Indian and Pākehā
  • There is separate study for Māori(running at the moment too)
  • This talk presents the sampling design and lessons learned from the IDEA study

Slide 1

# Load the DiagrammeR package
library(DiagrammeR)

# Create the flow diagram with rounded corners for the boxes
# Load the DiagrammeR package
library(DiagrammeR)

# Create the flow diagram with rounded corners, custom fill colors, and horizontal layout
graph <- "
digraph flow {

  # Set graph layout to horizontal (left to right)
  rankdir = LR;

  # Define node properties with rounded corners and fill colors
  node [shape = rect, style = filled, fontname = Helvetica, fontsize = 12, width = 2, height = 0.6, color = black, fontcolor = black, style = rounded];

  # Define the nodes (phases) with custom fill colors
  Areas [label = 'Phase 1: Areas', fillcolor = lightgreen, width = 2.5]
  Meshblocks [label = 'Phase 2: Meshblocks', fillcolor = lightblue, width = 2.5]
  Screening [label = 'Phase 3: Participants', fillcolor = lightcoral, width = 2.5]

  # Define the edges (arrows) with labels and colors
  Areas -> Meshblocks [label = '', fontsize = 10, color = blue, fontcolor = blue];
  Meshblocks -> Screening [label = '', fontsize = 10, color = green, fontcolor = green];

  # Styling the edges (arrows)
  edge [arrowhead = vee, arrowsize = 1.5, color = gray];
}
"

#grViz(graph)

# Render the diagram with horizontal layout

Sampling strategy

Frame: Territorial autorities

Image 1

Sampling plan

Image

  • Apart from the phases, our desing had three main features:

    • Stratification by TA and rurality
    • Oversampling of areas with high density of Chinese and Indian 65+
    • Proportion of meshblocks sampled from each area depends on the density of Chinese and Indian.
    • Same precision of prevalacen estimates for the three ethnicities.

  • Phase 1: Areas

  • Stratification by district (AKL or CHC) & rural/urban

  • Sample size allocation: 75% for AKL and 25% for CHC

  • 30%+ Chinese or 20%+ Indian all have a high chance of selection (60%+)

  • Phase 2: Meshblocks

  • Proportion of Meshblock selcted form each area

  • Denser Chinese/Indian meshblocks have higher chances of selection

  • Phase 3: Participants

  • Screening tool for dementia- everyone Positive selected, 50% negative selected

  • Only 30% pakeha selected

Sample sizes for each domain

  • We run lots of simulations to identify a sensible total sampl
  • Image

We decided on a margin of error of about 0.03,

Phase 1: Areas

Phase 1: Areas

## ℹ Google's Terms of Service: <https://mapsplatform.google.com>
##   Stadia Maps' Terms of Service: <https://stadiamaps.com/terms-of-service/>
##   OpenStreetMap's Tile Usage Policy: <https://operations.osmfoundation.org/policies/tiles/>
## ℹ Please cite ggmap if you use it! Use `citation("ggmap")` for details.
## 
## Attaching package: 'ggmap'
## 
## 
## The following object is masked from 'package:plotly':
## 
##     wind

Recruitment over time